Pueraria lobata (P. lobata), a traditional anti-diabetic medicine mainly composed of flavonoids and isoflavones, has a long history in diabetes treatment in China. However, the anti-diabetic active component is still unclear. Recently, protein tyrosine phosphatase 1B (PTP1B) has been a hot therapeutic target by negatively regulating insulin signaling pathways. In this study, the spectrum-effect relationship analysis method was first used to identify the active components of P. lobata that inhibit PTP1B. The fingerprints of 12 batches of samples were established using high-performance liquid chromatography (HPLC), and sixty common peaks were identified. Meanwhile, twelve components were identified by a comparison with the standards. The inhibition of PTP1B activity was studied in vitro by using the p-nitrophenol method, and the partial least squares discriminant analysis, grey relational analysis, bivariate correlation analysis, and cluster analysis were used to analyze the bioactive compounds in P. lobata. Peaks 6, 9 (glycitin), 11 (genistin), 12 (4'-methoxypuerarin), 25, 34, 35, 36, 53, and 59 were considered as potentially active substances that inhibit PTP1B. The in vitro PTP1B inhibitory activity was confirmed by glycitin, genistin, and 4'-methoxypuerarin. The IC50s of the three compounds were 10.56 ± 0.42 μg/mL, 16.46 ± 0.29 μg/mL, and 9.336 ± 0.56 μg/mL, respectively, indicating the obvious PTP1B inhibitory activity. In brief, we established an effective method to identify PTP1B enzyme inhibitors in P. lobata, which is helpful in clarifying the material basis of P. lobata on diabetes. Additionally, it is evident that the spectrum-effect relationship method serves as an efficient approach for identifying active compounds, and this study can also serve as a reference for screening bioactive constituents in traditional Chinese medicine.